An AFP report published March 21 details a proposal from Mistral AI’s CEO: impose a levy of 1.0 to 1.5 percent of revenues on AI model sales in Europe, direct the proceeds to a central European fund supporting content creation and cultural industries, and, critically, extend the levy equally to foreign AI providers.
This is a corporate proposal, not a regulatory announcement. No EU institution has moved on this.
The vendor interest here is real and worth naming. Mistral, which would benefit directly from the legal certainty the levy is intended to provide, argues that this mechanism would resolve the AI training and copyright standoff without litigation. That framing is Mistral’s own characterization, per the AFP report.
The proposal lands alongside active EU litigation. The CJEU held its first oral hearing on AI training and copyright on March 10 in *Like Company v. Google*, a case examining whether LLM training violates European copyright law. A levy like the one Mistral proposes would, in theory, moot cases like that one by replacing the litigation framework with a licensing framework.
Whether EU regulators or legislators find the proposal viable is unknown. The equal-application clause for foreign providers, meaning US companies like Google, OpenAI, and Anthropic would pay the same levy as European AI companies, is politically useful for Mistral’s framing but would face significant pushback in transatlantic trade discussions.
For compliance teams monitoring EU AI obligations, this is a proposal to track, not a requirement to plan for.